Fig 1.
The optimal classification surface.
Fig 2.
Framework of the AFWU-SVM algorithm.
Fig 3.
Processing flow of a task using the MapReduce model.
k, the key of the key-value pairs; v, the value of the key-value pairs.
Fig 4.
Overall framework of the PAFWU-SVM algorithm.
Fig 5.
Comparison of the speedup values.
Table 1.
Comparison of the classification accuracies of different algorithms and different feature fusion methods.
Fig 6.
Comparison of the average classification accuracies of the proposed algorithm for different numbers of classification categories with varying numbers of images (Notes: 10 categories: 1,200 images; 50 categories: 6,000 images; 150 categories: 16,500 images; 300 categories: 31,800 images; 500 categories: 51,095 images; 635 categories: 80,000 images).
Fig 7.
Comparison of the training times among different algorithms.
Table 2.
Comprehensive comparison between the algorithm proposed in this paper and other deep learning algorithms.